Research On Highway Tunnel Fire Detection Technology Based On Image Processing

Fire accident is a kind of accident which is very harmful in highway tunnel safety accident,and it often causes serious casualties and great property losses.If the fire can be detected accurately and timely in the early stage,the damage will be minimized.So,the early detection of road tunnel fires is extremely important.The traditional highway tunnel fire detection technology is affected by the complex environment of the tunnel,and it is difficult to realize timely and accurate detection.The image-based fire detection technology has characteristics of high accuracy,good real-time performance,strong anti-interference and low cost,which is suitable for highway tunnel fire detection.Combining with the complex environment of highway tunnel,taking fire flame as the research object and image processing as technical means,analyzing flame characteristics,studying the extraction method of flame candidate area and the extraction method of flame candidate area features,combining pattern recognition technology,this thesis puts forward a fire detection method suitable for highway tunnel.Firstly,the thesis analyzes characteristics of fire flame and studies the methods of eliminating common interferences in highway tunnels to design the fire detection method suitable for highway tunnel.Some theoretical knowledge of image processing in the process of fire detection is analyzed and studied,which provides research ideas and theoretical basis for the following research.Second,in order to reduce the influence of interference,this paper studies the extraction method of flame candidate area to extract accurate flame candidate regions.Firstly,the image is preprocessed with median filtering to reduce the influence of noise on the image.Secondly,the extraction method of moving target is studied in depth,and the ViBe background modeling difference method based on GMM is proposed to extract the moving target.Then,the color distribution of flame in the color space model is analyzed,and a method of segmentation of suspected flame region based on YCbCr color space and saturation is proposed.Finally,the moving target region and the suspected flame region are combined to determine the candidate region of the flame,and the extracted flame candidates are selected.The area is subjected to morphological processing and connected domain marking to prepare for the next step of extracting the characteristics of the flame.Then,according to the characteristics of the flame,the colour first moment,the circularity,the texture feature and the flicker characteristic feature of the flame candidate region are extracted,and used as the input vector of the support vector machine model to select the appropriate kernel function,optimize the kernel function parameters,and train the appropriate support vector machine classifier model.Finally,in order to verify the effectiveness and feasibility of the highway tunnel fire detection technology based on image processing proposed in this thesis,the technology is used to detect multiple flame video and interference video.The test results show that the technology has high flame recognition accuracy and interference removal rate.So,it can be applied to fire detection and identification of highway tunnels.